Author(s)

V. Anh, K. Lunney, P. Best, G. Johnson, M. Azzi & H. Duc

Abstract

This paper describes a fractional autoregressive model and a multivariate causality model for prediction of maximum daily Lidcombe ozone concentration. The models accommodate long-range dependence, which is an important aspect of concentration time series. It is found that morning wind speed, temperature, extent and ozone measured at 11 am contain information which can be used to improve the forecasts of univariate models which rely on the history of the maximum daily ozone series alone. With these factors incorporated, the resulting causality model gives a much improved performance on predicting ozone episodes. The

Keywords